
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import scipy

from sklearn.linear_model import SGDRegressor
from sklearn.preprocessing import StandardScaler
from sklearn.preprocessing import MinMaxScaler
from sklearn.pipeline import Pipeline
from sklearn.metrics import mean_absolute_error
from sklearn import linear_model
from sklearn.metrics import mean_squared_error, r2_score

from sklearn.decomposition import PCA
from sklearn.decomposition import TruncatedSVD
from sklearn.preprocessing import StandardScaler
from scipy.sparse import csr_matrix

import matplotlib.pyplot as plt
import seaborn as sns

from sklearn.model_selection import train_test_split

from sklearn.tree import DecisionTreeClassifier
from sklearn import tree
from sklearn.tree import export_text

from sklearn import metrics

import seaborn as sns
import matplotlib.pyplot as plt


heroes = pd.read_csv('archive/hero_names.csv')
heroes.head()


players = pd.read_csv('archive/players.csv')
players.fillna(0)
players = players[['match_id', 'account_id', 'hero_id', 'player_slot', 'gold', 'gold_per_min', 'xp_per_min', 'kills', 'deaths', 'assists', 'denies', 'hero_damage', 'hero_healing', 'level', 'leaver_status']]
players.head()


players['leave'] = 0
players.loc[players['leaver_status'] == 2, 'leave'] = 1
players.loc[players['leaver_status'] == 3, 'leave'] = 1
players.loc[players['leaver_status'] == 4, 'leave'] = 1


ratings = pd.read_csv('archive/player_ratings.csv')
ratings.head()


regions = pd.read_csv('archive/cluster_regions.csv')
regions.head()


matches = pd.read_csv('archive/match.csv')
matches = matches[['match_id', 'cluster']]
matches.head()


playerHeroes = pd.merge(players, heroes, how='inner', left_on = 'hero_id', right_on = 'hero_id')
playerHeroes.drop(columns='name', inplace=True)


df = pd.merge(playerHeroes, ratings, how='inner', left_on = 'account_id', right_on = 'account_id')
df = pd.merge(df, pd.merge(regions, matches, how='inner', left_on = 'cluster', right_on = 'cluster'), how='inner', left_on = 'match_id', right_on = 'match_id')
df['stay'] = 1
df.loc[df['account_id'] == 0, 'stay'] = 0
df['team'] = 0
df.loc[df['player_slot'] > 5, 'team'] = 1
bf = df[df['account_id'] != 0]
df.head()


df.isnull().any()

match_id           False
account_id         False
hero_id            False
player_slot        False
gold               False
gold_per_min       False
xp_per_min         False
kills              False
deaths             False
assists            False
denies             False
hero_damage        False
hero_healing       False
level              False
leaver_status      False
leave              False
localized_name     False
total_wins         False
total_matches      False
trueskill_mu       False
trueskill_sigma    False
cluster            False
region             False
stay               False
team               False
dtype: bool


df['trueskill_mu'].hist(bins=4)

<AxesSubplot:>


df['trueskill_sigma'].hist(bins=4)

<AxesSubplot:>


df.groupby('stay').mean()


df.groupby('leave').mean()


df.groupby('leaver_status').mean()


df.groupby('team').mean()


from scipy.stats import ttest_ind
print('gold: ', ttest_ind(bf['gold'], df[df['account_id'] == 0]['gold']))
print('gold_per_min: ', ttest_ind(bf['gold_per_min'], df[df['account_id'] == 0]['gold_per_min']))
print('xp_per_min: ', ttest_ind(bf['xp_per_min'], df[df['account_id'] == 0]['xp_per_min']))
print('kills: ', ttest_ind(bf['kills'], df[df['account_id'] == 0]['kills']))
print('deaths: ', ttest_ind(bf['deaths'], df[df['account_id'] == 0]['deaths']))
print('assists: ', ttest_ind(bf['assists'], df[df['account_id'] == 0]['assists']))
print('denies: ', ttest_ind(bf['denies'], df[df['account_id'] == 0]['denies']))
print('hero_damage: ', ttest_ind(bf['hero_damage'], df[df['account_id'] == 0]['hero_damage']))
print('hero_healing: ', ttest_ind(bf['hero_healing'], df[df['account_id'] == 0]['hero_healing']))

gold:  Ttest_indResult(statistic=29.094556485437266, pvalue=6.182275805081904e-186)
gold_per_min:  Ttest_indResult(statistic=32.79333413095849, pvalue=1.3634458787213026e-235)
xp_per_min:  Ttest_indResult(statistic=10.59672013822142, pvalue=3.1074674444333967e-26)
kills:  Ttest_indResult(statistic=-2.201987146900172, pvalue=0.02766670462726259)
deaths:  Ttest_indResult(statistic=-52.25797798410857, pvalue=0.0)
assists:  Ttest_indResult(statistic=8.913550755542275, pvalue=4.959566898847809e-19)
denies:  Ttest_indResult(statistic=68.31598030982597, pvalue=0.0)
hero_damage:  Ttest_indResult(statistic=2.04719043512754, pvalue=0.04063994994516191)
hero_healing:  Ttest_indResult(statistic=17.19153047290371, pvalue=3.221561896486699e-66)


from scipy.stats import ttest_ind
print('gold: ', ttest_ind(df[df['leave'] == 0]['gold'], df[df['leave'] == 1]['gold']))
print('gold_per_min: ', ttest_ind(df[df['leave'] == 0]['gold_per_min'], df[df['leave'] == 1]['gold_per_min']))
print('xp_per_min: ', ttest_ind(df[df['leave'] == 0]['xp_per_min'], df[df['leave'] == 1]['xp_per_min']))
print('kills: ', ttest_ind(df[df['leave'] == 0]['kills'], df[df['leave'] == 1]['kills']))
print('deaths: ', ttest_ind(df[df['leave'] == 0]['deaths'], df[df['leave'] == 1]['deaths']))
print('assists: ', ttest_ind(df[df['leave'] == 0]['assists'], df[df['leave'] == 1]['assists']))
print('denies: ', ttest_ind(df[df['leave'] == 0]['denies'], df[df['leave'] == 1]['denies']))
print('hero_damage: ', ttest_ind(df[df['leave'] == 0]['hero_damage'], df[df['leave'] == 1]['hero_damage']))
print('hero_healing: ', ttest_ind(df[df['leave'] == 0]['hero_healing'], df[df['leave'] == 1]['hero_healing']))

gold:  Ttest_indResult(statistic=58.43287601218118, pvalue=0.0)
gold_per_min:  Ttest_indResult(statistic=74.31418083832251, pvalue=0.0)
xp_per_min:  Ttest_indResult(statistic=95.92505442985849, pvalue=0.0)
kills:  Ttest_indResult(statistic=50.45322516837418, pvalue=0.0)
deaths:  Ttest_indResult(statistic=13.257583958164963, pvalue=4.147180207023009e-40)
assists:  Ttest_indResult(statistic=59.64536775311154, pvalue=0.0)
denies:  Ttest_indResult(statistic=20.360381990924903, pvalue=4.120954087520137e-92)
hero_damage:  Ttest_indResult(statistic=58.425296268144734, pvalue=0.0)
hero_healing:  Ttest_indResult(statistic=17.06008695047411, pvalue=3.0788628981518285e-65)


df['mu_bin'] = pd.cut(df['trueskill_mu'], 4, labels=[1,2,3,4])
df['sigma_bin'] = pd.cut(df['trueskill_mu'], 4, labels=[1,2,3,4])


def apriori(dataset, min_support=0.5, verbose=False):
    """Implements the Apriori algorithm.

    The Apriori algorithm will iteratively generate new candidate 
    k-itemsets using the frequent (k-1)-itemsets found in the previous 
    iteration.

    Parameters
    ----------
    dataset : list
        The dataset (a list of transactions) from which to generate 
        candidate itemsets.

    min_support : float
        The minimum support threshold. Defaults to 0.5.

    Returns
    -------
    F : list
        The list of frequent itemsets.

    support_data : dict
        The support data for all candidate itemsets.

    References
    ----------
    .. [1] R. Agrawal, R. Srikant, "Fast Algorithms for Mining Association 
           Rules", 1994.

    """
    C1 = create_candidates(dataset)
    D = list(map(set, dataset))
    F1, support_data = support_prune(D, C1, min_support, verbose=False) # prune candidate 1-itemsets
    F = [F1] # list of frequent itemsets; initialized to frequent 1-itemsets
    k = 2 # the itemset cardinality
    while (len(F[k - 2]) > 0):
        Ck = apriori_gen(F[k-2], k) # generate candidate itemsets
        Fk, supK = support_prune(D, Ck, min_support) # prune candidate itemsets
        support_data.update(supK) # update the support counts to reflect pruning
        F.append(Fk) # add the pruned candidate itemsets to the list of frequent itemsets
        k += 1

    if verbose:
        # Print a list of all the frequent itemsets.
        for kset in F:
            for item in kset:
                print("" \
                    + "{" \
                    + "".join(str(i) + ", " for i in iter(item)).rstrip(', ') \
                    + "}" \
                    + ":  sup = " + str(round(support_data[item], 3)))

    return F, support_data

def create_candidates(dataset, verbose=False):
    """Creates a list of candidate 1-itemsets from a list of transactions.

    Parameters
    ----------
    dataset : list
        The dataset (a list of transactions) from which to generate candidate 
        itemsets.

    Returns
    -------
    The list of candidate itemsets (c1) passed as a frozenset (a set that is 
    immutable and hashable).
    """
    c1 = [] # list of all items in the database of transactions
    for transaction in dataset:
        for item in transaction:
            if not [item] in c1:
                c1.append([item])
    c1.sort()

    if verbose:
        # Print a list of all the candidate items.
        print("" \
            + "{" \
            + "".join(str(i[0]) + ", " for i in iter(c1)).rstrip(', ') \
            + "}")

    # Map c1 to a frozenset because it will be the key of a dictionary.
    return list(map(frozenset, c1))

def support_prune(dataset, candidates, min_support, verbose=False):
    """Returns all candidate itemsets that meet a minimum support threshold.

    By the apriori principle, if an itemset is frequent, then all of its 
    subsets must also be frequent. As a result, we can perform support-based 
    pruning to systematically control the exponential growth of candidate 
    itemsets. Thus, itemsets that do not meet the minimum support level are 
    pruned from the input list of itemsets (dataset).

    Parameters
    ----------
    dataset : list
        The dataset (a list of transactions) from which to generate candidate 
        itemsets.

    candidates : frozenset
        The list of candidate itemsets.

    min_support : float
        The minimum support threshold.

    Returns
    -------
    retlist : list
        The list of frequent itemsets.

    support_data : dict
        The support data for all candidate itemsets.
    """
    sscnt = {} # set for support counts
    for tid in dataset:
        for can in candidates:
            if can.issubset(tid):
                sscnt.setdefault(can, 0)
                sscnt[can] += 1

    num_items = float(len(dataset)) # total number of transactions in the dataset
    retlist = [] # array for unpruned itemsets
    support_data = {} # set for support data for corresponding itemsets
    for key in sscnt:
        # Calculate the support of itemset key.
        support = sscnt[key] / num_items
        if support >= min_support:
            retlist.insert(0, key)
        support_data[key] = support

    # Print a list of the pruned itemsets.
    if verbose:
        for kset in retlist:
            for item in kset:
                print("{" + str(item) + "}")
        print("")
        for key in sscnt:
            print("" \
                + "{" \
                + "".join([str(i) + ", " for i in iter(key)]).rstrip(', ') \
                + "}" \
                + ":  sup = " + str(support_data[key]))

    return retlist, support_data

def apriori_gen(freq_sets, k):
    """Generates candidate itemsets (via the F_k-1 x F_k-1 method).

    This operation generates new candidate k-itemsets based on the frequent 
    (k-1)-itemsets found in the previous iteration. The candidate generation 
    procedure merges a pair of frequent (k-1)-itemsets only if their first k-2 
    items are identical.

    Parameters
    ----------
    freq_sets : list
        The list of frequent (k-1)-itemsets.

    k : integer
        The cardinality of the current itemsets being evaluated.

    Returns
    -------
    retlist : list
        The list of merged frequent itemsets.
    """
    retList = [] # list of merged frequent itemsets
    lenLk = len(freq_sets) # number of frequent itemsets
    for i in range(lenLk):
        for j in range(i+1, lenLk):
            a=list(freq_sets[i])
            b=list(freq_sets[j])
            a.sort()
            b.sort()
            F1 = a[:k-2] # first k-2 items of freq_sets[i]
            F2 = b[:k-2] # first k-2 items of freq_sets[j]

            if F1 == F2: # if the first k-2 items are identical
                # Merge the frequent itemsets.
                retList.append(freq_sets[i] | freq_sets[j])

    return retList

def rules_from_conseq(freq_set, H, support_data, rules, min_confidence=0.5, verbose=False):
    """Generates a set of candidate rules.

    Parameters
    ----------
    freq_set : frozenset
        The complete list of frequent itemsets.

    H : list
        A list of frequent itemsets (of a particular length).

    support_data : dict
        The support data for all candidate itemsets.

    rules : list
        A potentially incomplete set of candidate rules above the minimum 
        confidence threshold.

    min_confidence : float
        The minimum confidence threshold. Defaults to 0.5.
    """
    m = len(H[0])
    if m == 1:
        Hmp1 = calc_confidence(freq_set, H, support_data, rules, min_confidence, verbose)
    if (len(freq_set) > (m+1)):
        Hmp1 = apriori_gen(H, m+1) # generate candidate itemsets
        Hmp1 = calc_confidence(freq_set, Hmp1, support_data, rules, min_confidence, verbose)
        if len(Hmp1) > 1:
            # If there are candidate rules above the minimum confidence 
            # threshold, recurse on the list of these candidate rules.
            rules_from_conseq(freq_set, Hmp1, support_data, rules, min_confidence, verbose)

def calc_confidence(freq_set, H, support_data, rules, min_confidence=0.5, verbose=False):
    """Evaluates the generated rules.

    One measurement for quantifying the goodness of association rules is 
    confidence. The confidence for a rule 'P implies H' (P -> H) is defined as 
    the support for P and H divided by the support for P 
    (support (P|H) / support(P)), where the | symbol denotes the set union 
    (thus P|H means all the items in set P or in set H).

    To calculate the confidence, we iterate through the frequent itemsets and 
    associated support data. For each frequent itemset, we divide the support 
    of the itemset by the support of the antecedent (left-hand-side of the 
    rule).

    Parameters
    ----------
    freq_set : frozenset
        The complete list of frequent itemsets.

    H : list
        A list of frequent itemsets (of a particular length).

    min_support : float
        The minimum support threshold.

    rules : list
        A potentially incomplete set of candidate rules above the minimum 
        confidence threshold.

    min_confidence : float
        The minimum confidence threshold. Defaults to 0.5.

    Returns
    -------
    pruned_H : list
        The list of candidate rules above the minimum confidence threshold.
    """
    pruned_H = [] # list of candidate rules above the minimum confidence threshold
    for conseq in H: # iterate over the frequent itemsets
        conf = support_data[freq_set] / support_data[freq_set - conseq]
        if conf >= min_confidence:
            rules.append((freq_set - conseq, conseq, conf))
            pruned_H.append(conseq)

            if verbose:
                print("" \
                    + "{" \
                    + "".join([str(i) + ", " for i in iter(freq_set-conseq)]).rstrip(', ') \
                    + "}" \
                    + " ---> " \
                    + "{" \
                    + "".join([str(i) + ", " for i in iter(conseq)]).rstrip(', ') \
                    + "}" \
                    + ":  conf = " + str(round(conf, 3)) \
                    + ", sup = " + str(round(support_data[freq_set], 3)))

    return pruned_H

def generate_rules(F, support_data, min_confidence=0.5, verbose=True):
    """Generates a set of candidate rules from a list of frequent itemsets.

    For each frequent itemset, we calculate the confidence of using a
    particular item as the rule consequent (right-hand-side of the rule). By 
    testing and merging the remaining rules, we recursively create a list of 
    pruned rules.

    Parameters
    ----------
    F : list
        A list of frequent itemsets.

    support_data : dict
        The corresponding support data for the frequent itemsets (L).

    min_confidence : float
        The minimum confidence threshold. Defaults to 0.5.

    Returns
    -------
    rules : list
        The list of candidate rules above the minimum confidence threshold.
    """
    rules = []
    for i in range(1, len(F)):
        for freq_set in F[i]:
            H1 = [frozenset([itemset]) for itemset in freq_set]
            if (i > 1):
                rules_from_conseq(freq_set, H1, support_data, rules, min_confidence, verbose)
            else:
                calc_confidence(freq_set, H1, support_data, rules, min_confidence, verbose)

    return rules


testA = df[['player_slot', 'match_id', 'localized_name']]
testB = testA.copy()
testB['team'] = 0
testB.loc[testB['player_slot'] > 5, 'team'] = 1
testB.loc[testB['player_slot'] == 128, 'player_slot'] = 0
testB.loc[testB['player_slot'] == 129, 'player_slot'] = 1
testB.loc[testB['player_slot'] == 130, 'player_slot'] = 2
testB.loc[testB['player_slot'] == 131, 'player_slot'] = 3
testB.loc[testB['player_slot'] == 132, 'player_slot'] = 4


testA.set_index(['match_id','player_slot'])['localized_name'].unstack()
testA = testA.pivot(index='match_id', columns='player_slot', values='localized_name')


testB.set_index(['match_id', 'team','player_slot'])['localized_name'].unstack()
testB = testB.pivot(index=['match_id','team'], columns='player_slot', values='localized_name')


testA.dropna(inplace=True)
len(testA)

28738


testB.dropna(inplace=True)
len(testB)

74672


d = []
dl = []
for i in range(len(testA)):
    dl.append(testA.iloc[i,:].tolist())
    d.append(set(testA.iloc[i,:].tolist()))


D = d


C1 = create_candidates(np.array(dl), verbose=False)


F1, support_data = support_prune(D, C1, 0.03, verbose=False)


F, support_data = apriori(np.array(dl), min_support=0.03, verbose=False)
# sorts the dictionary by values and returns top 10
# list({k: v for k, v in sorted(support_data.items(), reverse=True, key=lambda item: item[1])}.items())[:10]


from collections import defaultdict, namedtuple
#from itertools import imap

__author__ = 'Eric Naeseth <eric@naeseth.com>'
__copyright__ = 'Copyright © 2009 Eric Naeseth'
__license__ = 'MIT License'

def fpgrowth(dataset, min_support=0.5, include_support=True, verbose=False):
    """Implements the FP-growth algorithm.

    The `dataset` parameter can be any iterable of iterables of items.
    `min_support` should be an integer specifying the minimum number of
    occurrences of an itemset for it to be accepted.

    Each item must be hashable (i.e., it must be valid as a member of a
    dictionary or a set).

    If `include_support` is true, yield (itemset, support) pairs instead of
    just the itemsets.

    Parameters
    ----------
    dataset : list
        The dataset (a list of transactions) from which to generate 
        candidate itemsets.

    min_support : float
        The minimum support threshold. Defaults to 0.5.

    include_support : bool
        Include support in output (default=False).

    References
    ----------
    .. [1] J. Han, J. Pei, Y. Yin, "Mining Frequent Patterns without Candidate 
           Generation," 2000.

    """

    F = []
    support_data = {}
    for k,v in find_frequent_itemsets(dataset, min_support=min_support, include_support=include_support, verbose=verbose):
        F.append(frozenset(k))
        support_data[frozenset(k)] = v

    # Create one array with subarrays that hold all transactions of equal length.
    def bucket_list(nested_list, sort=True):
        bucket = defaultdict(list)
        for sublist in nested_list:
            bucket[len(sublist)].append(sublist)
        return [v for k,v in sorted(bucket.items())] if sort else bucket.values()

    F = bucket_list(F)
    
    return F, support_data

def find_frequent_itemsets(dataset, min_support, include_support=False, verbose=False):
    """
    Find frequent itemsets in the given transactions using FP-growth. This
    function returns a generator instead of an eagerly-populated list of items.

    The `dataset` parameter can be any iterable of iterables of items.
    `min_support` should be an integer specifying the minimum number of
    occurrences of an itemset for it to be accepted.

    Each item must be hashable (i.e., it must be valid as a member of a
    dictionary or a set).

    If `include_support` is true, yield (itemset, support) pairs instead of
    just the itemsets.

    Parameters
    ----------
    dataset : list
        The dataset (a list of transactions) from which to generate 
        candidate itemsets.

    min_support : float
        The minimum support threshold. Defaults to 0.5.

    include_support : bool
        Include support in output (default=False).

    """
    items = defaultdict(lambda: 0) # mapping from items to their supports
    processed_transactions = []

    # Load the passed-in transactions and count the support that individual
    # items have.
    for transaction in dataset:
        processed = []
        for item in transaction:
            items[item] += 1
            processed.append(item)
        processed_transactions.append(processed)

    # Remove infrequent items from the item support dictionary.
    items = dict((item, support) for item, support in items.items()
        if support >= min_support)

    # Build our FP-tree. Before any transactions can be added to the tree, they
    # must be stripped of infrequent items and their surviving items must be
    # sorted in decreasing order of frequency.
    def clean_transaction(transaction):
        #transaction = filter(lambda v: v in items, transaction)
        transaction.sort(key=lambda v: items[v], reverse=True)
        return transaction

    master = FPTree()
    for transaction in map(clean_transaction, processed_transactions):
        master.add(transaction)

    support_data = {}
    def find_with_suffix(tree, suffix):
        for item, nodes in tree.items():
            support = float(sum(n.count for n in nodes)) / len(dataset)
            if support >= min_support and item not in suffix:
                # New winner!
                found_set = [item] + suffix
                support_data[frozenset(found_set)] = support
                yield (found_set, support) if include_support else found_set

                # Build a conditional tree and recursively search for frequent
                # itemsets within it.
                cond_tree = conditional_tree_from_paths(tree.prefix_paths(item),
                    min_support)
                for s in find_with_suffix(cond_tree, found_set):
                    yield s # pass along the good news to our caller

    if verbose:
        # Print a list of all the frequent itemsets.
        for itemset, support in find_with_suffix(master, []):
            print("" \
                + "{" \
                + "".join(str(i) + ", " for i in iter(itemset)).rstrip(', ') \
                + "}" \
                + ":  sup = " + str(round(support_data[frozenset(itemset)], 3)))

    # Search for frequent itemsets, and yield the results we find.
    for itemset in find_with_suffix(master, []):
        yield itemset

class FPTree(object):
    """
    An FP tree.

    This object may only store transaction items that are hashable (i.e., all
    items must be valid as dictionary keys or set members).
    """

    Route = namedtuple('Route', 'head tail')

    def __init__(self):
        # The root node of the tree.
        self._root = FPNode(self, None, None)

        # A dictionary mapping items to the head and tail of a path of
        # "neighbors" that will hit every node containing that item.
        self._routes = {}

    @property
    def root(self):
        """The root node of the tree."""
        return self._root

    def add(self, transaction):
        """
        Adds a transaction to the tree.
        """

        point = self._root

        for item in transaction:
            next_point = point.search(item)
            if next_point:
                # There is already a node in this tree for the current
                # transaction item; reuse it.
                next_point.increment()
            else:
                # Create a new point and add it as a child of the point we're
                # currently looking at.
                next_point = FPNode(self, item)
                point.add(next_point)

                # Update the route of nodes that contain this item to include
                # our new node.
                self._update_route(next_point)

            point = next_point

    def _update_route(self, point):
        """Add the given node to the route through all nodes for its item."""
        assert self is point.tree

        try:
            route = self._routes[point.item]
            route[1].neighbor = point # route[1] is the tail
            self._routes[point.item] = self.Route(route[0], point)
        except KeyError:
            # First node for this item; start a new route.
            self._routes[point.item] = self.Route(point, point)

    def items(self):
        """
        Generate one 2-tuples for each item represented in the tree. The first
        element of the tuple is the item itself, and the second element is a
        generator that will yield the nodes in the tree that belong to the item.
        """
        for item in self._routes.keys():
            yield (item, self.nodes(item))

            
    def nodes(self, item):
        """
        Generates the sequence of nodes that contain the given item.
        """

        try:
            node = self._routes[item][0]
        except KeyError:
            return

        while node:
            yield node
            node = node.neighbor

    def prefix_paths(self, item):
        """Generates the prefix paths that end with the given item."""

        def collect_path(node):
            path = []
            while node and not node.root:
                path.append(node)
                node = node.parent
            path.reverse()
            return path

        return (collect_path(node) for node in self.nodes(item))

    def inspect(self):
        print("Tree:")
        self.root.inspect(1)

        print("")
        print("Routes:")
        for item, nodes in self.items():
            print("  %r" % item)
            for node in nodes:
                print("    %r" % node)

    def _removed(self, node):
        """Called when `node` is removed from the tree; performs cleanup."""

        head, tail = self._routes[node.item]
        if node is head:
            if node is tail or not node.neighbor:
                # It was the sole node.
                del self._routes[node.item]
            else:
                self._routes[node.item] = self.Route(node.neighbor, tail)
        else:
            for n in self.nodes(node.item):
                if n.neighbor is node:
                    n.neighbor = node.neighbor # skip over
                    if node is tail:
                        self._routes[node.item] = self.Route(head, n)
                    break

def conditional_tree_from_paths(paths, min_support):
    """Builds a conditional FP-tree from the given prefix paths."""
    tree = FPTree()
    condition_item = None
    items = set()

    # Import the nodes in the paths into the new tree. Only the counts of the
    # leaf notes matter; the remaining counts will be reconstructed from the
    # leaf counts.
    for path in paths:
        if condition_item is None:
            condition_item = path[-1].item

        point = tree.root
        for node in path:
            next_point = point.search(node.item)
            if not next_point:
                # Add a new node to the tree.
                items.add(node.item)
                count = node.count if node.item == condition_item else 0
                next_point = FPNode(tree, node.item, count)
                point.add(next_point)
                tree._update_route(next_point)
            point = next_point

    assert condition_item is not None

    # Calculate the counts of the non-leaf nodes.
    for path in tree.prefix_paths(condition_item):
        count = path[-1].count
        for node in reversed(path[:-1]):
            node._count += count

    # Eliminate the nodes for any items that are no longer frequent.
    for item in items:
        support = sum(n.count for n in tree.nodes(item))
        if support < min_support:
            # Doesn't make the cut anymore
            for node in tree.nodes(item):
                if node.parent is not None:
                    node.parent.remove(node)

    # Finally, remove the nodes corresponding to the item for which this
    # conditional tree was generated.
    for node in tree.nodes(condition_item):
        if node.parent is not None: # the node might already be an orphan
            node.parent.remove(node)

    return tree

class FPNode(object):
    """A node in an FP tree."""

    def __init__(self, tree, item, count=1):
        self._tree = tree
        self._item = item
        self._count = count
        self._parent = None
        self._children = {}
        self._neighbor = None

    def add(self, child):
        """Adds the given FPNode `child` as a child of this node."""

        if not isinstance(child, FPNode):
            raise TypeError("Can only add other FPNodes as children")

        if not child.item in self._children:
            self._children[child.item] = child
            child.parent = self

    def search(self, item):
        """
        Checks to see if this node contains a child node for the given item.
        If so, that node is returned; otherwise, `None` is returned.
        """

        try:
            return self._children[item]
        except KeyError:
            return None

    def remove(self, child):
        try:
            if self._children[child.item] is child:
                del self._children[child.item]
                child.parent = None
                self._tree._removed(child)
                for sub_child in child.children:
                    try:
                        # Merger case: we already have a child for that item, so
                        # add the sub-child's count to our child's count.
                        self._children[sub_child.item]._count += sub_child.count
                        sub_child.parent = None # it's an orphan now
                    except KeyError:
                        # Turns out we don't actually have a child, so just add
                        # the sub-child as our own child.
                        self.add(sub_child)
                child._children = {}
            else:
                raise ValueError("that node is not a child of this node")
        except KeyError:
            raise ValueError("that node is not a child of this node")

    def __contains__(self, item):
        return item in self._children

    @property
    def tree(self):
        """The tree in which this node appears."""
        return self._tree

    @property
    def item(self):
        """The item contained in this node."""
        return self._item

    @property
    def count(self):
        """The count associated with this node's item."""
        return self._count

    def increment(self):
        """Increments the count associated with this node's item."""
        if self._count is None:
            raise ValueError("Root nodes have no associated count.")
        self._count += 1

    @property
    def root(self):
        """True if this node is the root of a tree; false if otherwise."""
        return self._item is None and self._count is None

    @property
    def leaf(self):
        """True if this node is a leaf in the tree; false if otherwise."""
        return len(self._children) == 0

    def parent():
        doc = "The node's parent."
        def fget(self):
            return self._parent
        def fset(self, value):
            if value is not None and not isinstance(value, FPNode):
                raise TypeError("A node must have an FPNode as a parent.")
            if value and value.tree is not self.tree:
                raise ValueError("Cannot have a parent from another tree.")
            self._parent = value
        return locals()
    parent = property(**parent())

    def neighbor():
        doc = """
        The node's neighbor; the one with the same value that is "to the right"
        of it in the tree.
        """
        def fget(self):
            return self._neighbor
        def fset(self, value):
            if value is not None and not isinstance(value, FPNode):
                raise TypeError("A node must have an FPNode as a neighbor.")
            if value and value.tree is not self.tree:
                raise ValueError("Cannot have a neighbor from another tree.")
            self._neighbor = value
        return locals()
    neighbor = property(**neighbor())

    @property
    def children(self):
        """The nodes that are children of this node."""
        return tuple(self._children.values())
        
    def inspect(self, depth=0):
        print(('  ' * depth) + repr(self))
        for child in self.children:
            child.inspect(depth + 1)

    def __repr__(self):
        if self.root:
            return "<%s (root)>" % type(self).__name__
        return "<%s %r (%r)>" % (type(self).__name__, self.item, self.count)

def rules_from_conseq(freq_set, H, support_data, rules, min_confidence=0.5, verbose=False):
    """Generates a set of candidate rules.

    Parameters
    ----------
    freq_set : frozenset
        The complete list of frequent itemsets.

    H : list
        A list of frequent itemsets (of a particular length).

    support_data : dict
        The support data for all candidate itemsets.

    rules : list
        A potentially incomplete set of candidate rules above the minimum 
        confidence threshold.

    min_confidence : float
        The minimum confidence threshold. Defaults to 0.5.
    """
    m = len(H[0])
    if m == 1:
        Hmp1 = calc_confidence(freq_set, H, support_data, rules, min_confidence, verbose)
    if (len(freq_set) > (m+1)):
        Hmp1 = apriori_gen(H, m+1) # generate candidate itemsets
        Hmp1 = calc_confidence(freq_set, Hmp1,  support_data, rules, min_confidence, verbose)
        if len(Hmp1) > 1:
            # If there are candidate rules above the minimum confidence 
            # threshold, recurse on the list of these candidate rules.
            rules_from_conseq(freq_set, Hmp1, support_data, rules, min_confidence, verbose)

def calc_confidence(freq_set, H, support_data, rules, min_confidence=0.5, verbose=False):
    """Evaluates the generated rules.

    One measurement for quantifying the goodness of association rules is 
    confidence. The confidence for a rule 'P implies H' (P -> H) is defined as 
    the support for P and H divided by the support for P 
    (support (P|H) / support(P)), where the | symbol denotes the set union 
    (thus P|H means all the items in set P or in set H).

    To calculate the confidence, we iterate through the frequent itemsets and 
    associated support data. For each frequent itemset, we divide the support 
    of the itemset by the support of the antecedent (left-hand-side of the 
    rule).

    Parameters
    ----------
    freq_set : frozenset
        The complete list of frequent itemsets.

    H : list
        A list of frequent itemsets (of a particular length).

    min_support : float
        The minimum support threshold.

    rules : list
        A potentially incomplete set of candidate rules above the minimum 
        confidence threshold.

    min_confidence : float
        The minimum confidence threshold. Defaults to 0.5.

    Returns
    -------
    pruned_H : list
        The list of candidate rules above the minimum confidence threshold.
    """
    pruned_H = [] # list of candidate rules above the minimum confidence threshold
    for conseq in H: # iterate over the frequent itemsets
        conf = support_data[freq_set] / support_data[freq_set - conseq]
        if conf >= min_confidence:
            rules.append((freq_set - conseq, conseq, conf))
            pruned_H.append(conseq)

            if verbose:
                print("" \
                    + "{" \
                    + "".join([str(i) + ", " for i in iter(freq_set-conseq)]).rstrip(', ') \
                    + "}" \
                    + " ---> " \
                    + "{" \
                    + "".join([str(i) + ", " for i in iter(conseq)]).rstrip(', ') \
                    + "}" \
                    + ":  conf = " + str(round(conf, 3)) \
                    + ", sup = " + str(round(support_data[freq_set], 3)))

    return pruned_H

def generate_rules(F, support_data, min_confidence=0.5, verbose=True):
    """Generates a set of candidate rules from a list of frequent itemsets.

    For each frequent itemset, we calculate the confidence of using a
    particular item as the rule consequent (right-hand-side of the rule). By 
    testing and merging the remaining rules, we recursively create a list of 
    pruned rules.

    Parameters
    ----------
    F : list
        A list of frequent itemsets.

    support_data : dict
        The corresponding support data for the frequent itemsets (L).

    min_confidence : float
        The minimum confidence threshold. Defaults to 0.5.

    Returns
    -------
    rules : list
        The list of candidate rules above the minimum confidence threshold.
    """
    rules = []
    for i in range(1, len(F)):
        for freq_set in F[i]:
            H1 = [frozenset([item]) for item in freq_set]
            if (i > 1):
                rules_from_conseq(freq_set, H1, support_data, rules, min_confidence, verbose)
            else:
                calc_confidence(freq_set, H1, support_data, rules, min_confidence, verbose)

    return rules


F, support_data = fpgrowth(np.array(dl), min_support=0.01, verbose=False)


H = generate_rules(F, support_data, min_confidence=0.2, verbose=False)


def Sort_Tuple(tup):  
        
    lst = len(tup)  
    for i in range(0, lst):  
            
        for j in range(0, lst-i-1):  
            if (tup[j][-1] > tup[j + 1][-1]):  
                temp = tup[j]  
                tup[j]= tup[j + 1]  
                tup[j + 1]= temp  
    return tup
          
confH = Sort_Tuple(H)[::-1]

print(confH[:5])

[(frozenset({'Slardar', 'Spectre'}), frozenset({'Windranger'}), 0.4416107382550336), (frozenset({'Slardar', 'Lich'}), frozenset({'Windranger'}), 0.4413875598086124), (frozenset({'Shadow Demon'}), frozenset({'Windranger'}), 0.44027303754266206), (frozenset({'Anti-Mage', 'Winter Wyvern'}), frozenset({'Windranger'}), 0.43654266958424504), (frozenset({'Bounty Hunter', 'Anti-Mage'}), frozenset({'Windranger'}), 0.4361233480176212)]


i = 1
for x in confH:
    print(x)
    if i >= 5:
        break
    i += 1

(frozenset({'Slardar', 'Spectre'}), frozenset({'Windranger'}), 0.4416107382550336)
(frozenset({'Slardar', 'Lich'}), frozenset({'Windranger'}), 0.4413875598086124)
(frozenset({'Shadow Demon'}), frozenset({'Windranger'}), 0.44027303754266206)
(frozenset({'Anti-Mage', 'Winter Wyvern'}), frozenset({'Windranger'}), 0.43654266958424504)
(frozenset({'Bounty Hunter', 'Anti-Mage'}), frozenset({'Windranger'}), 0.4361233480176212)


support = []
dups = []
for x in H:
    lst = list(x)
    l = set()
    for y in lst[:-1]:
        l.add(list(y)[0])

    # print(l, support_data[frozenset(l)])
    if (l, support_data[frozenset(l)]) not in support:
        support.append((l, support_data[frozenset(l)]))
    else:
        dups.append((l, support_data[frozenset(l)]))
support.sort(key=lambda x:x[1], reverse=True)


support[:5]

[({'Shadow Fiend', 'Windranger'}, 0.1312547846057485),
 ({'Slardar', 'Windranger'}, 0.09875426264875774),
 ({'Earthshaker', 'Windranger'}, 0.09008977660240797),
 ({'Shadow Fiend', 'Slardar'}, 0.08765397731226947),
 ({'Tusk', 'Windranger'}, 0.08744519451597188)]


from sklearn.model_selection import train_test_split

from sklearn.tree import DecisionTreeClassifier
from sklearn import tree
from sklearn.tree import export_text

from sklearn import metrics

import seaborn as sns
import matplotlib.pyplot as plt


x = df.drop(columns=['match_id', 'account_id', 'hero_id', 'player_slot', 'localized_name','cluster', 'region', 'team', 'stay', 'total_wins', 'total_matches', 'trueskill_mu', 'trueskill_sigma', 'mu_bin', 'sigma_bin', 'leaver_status', 'leave'])
y = df["mu_bin"]


X_train, test_x, y_train, test_y = train_test_split(x,y,
                                                 test_size = 0.3,
                                                 random_state = 42)


clf = DecisionTreeClassifier(random_state = 42)
clf.fit(X_train, y_train)

print('Max Depth: ', clf.get_depth())
print('Training Score: ', clf.score(X_train, y_train), 'Test Score: ', clf.score(test_x, test_y))
print('Number of Leaves: ', clf.get_n_leaves(), 'Feature Importances: ')
print(clf.feature_importances_)
print()

Max Depth:  67
Training Score:  1.0 Test Score:  0.6217138686549222
Number of Leaves:  70768 Feature Importances: 
[0.1700444  0.14474814 0.14961509 0.07151634 0.06057221 0.08309011
 0.0521177  0.16530853 0.05343677 0.04955071]


clf = DecisionTreeClassifier(random_state = 42, max_depth=5)
clf.fit(X_train, y_train)

print('Max Depth: ', clf.get_depth())
print('Training Score: ', clf.score(X_train, y_train), 'Test Score: ', clf.score(test_x, test_y))
print('Number of Leaves: ', clf.get_n_leaves(), 'Feature Importances: ')
print(clf.feature_importances_)
print()

Max Depth:  5
Training Score:  0.7588181176975681 Test Score:  0.7576670435193268
Number of Leaves:  32 Feature Importances: 
[0.00472083 0.17860602 0.03238515 0.00829522 0.17961894 0.01299292
 0.37670895 0.02106915 0.09639334 0.0892095 ]


x.head()


# for i in reversed(range(1, 40)):
#     clf = DecisionTreeClassifier(max_depth = i, random_state = 42)
#     clf.fit(X_train, y_train)
#     train = clf.score(X_train, y_train)
#     test = clf.score(test_x, test_y)
#     print('Max Depth: ', clf.get_depth())
#     print('Training Score: ', train, 'Test Score: ', test)
#     print('Number of Leaves: ', clf.get_n_leaves(), 'Feature Importances: ')
#     print(clf.feature_importances_)
#     print()


x = df.drop(columns=['match_id', 'account_id', 'hero_id', 'player_slot', 'localized_name','cluster', 'region', 'team', 'stay', 'total_wins', 'total_matches', 'trueskill_mu', 'trueskill_sigma', 'mu_bin', 'sigma_bin', 'leaver_status'])
y = df["sigma_bin"]


X_train, test_x, y_train, test_y = train_test_split(x,y,
                                                 test_size = 0.3,
                                                 random_state = 42)


clf = DecisionTreeClassifier(random_state = 42)
clf.fit(X_train, y_train)

print('Max Depth: ', clf.get_depth())
print('Training Score: ', clf.score(X_train, y_train), 'Test Score: ', clf.score(test_x, test_y))
print('Number of Leaves: ', clf.get_n_leaves(), 'Feature Importances: ')
print(clf.feature_importances_)
print()

Max Depth:  64
Training Score:  1.0 Test Score:  0.6222290596333557
Number of Leaves:  70701 Feature Importances: 
[0.16980806 0.14419618 0.14781792 0.0712137  0.05936863 0.08257973
 0.05282247 0.16828669 0.05278571 0.04997023 0.00115068]


clf = DecisionTreeClassifier(random_state = 42, max_depth=5)
clf.fit(X_train, y_train)

print('Max Depth: ', clf.get_depth())
print('Training Score: ', clf.score(X_train, y_train), 'Test Score: ', clf.score(test_x, test_y))
print('Number of Leaves: ', clf.get_n_leaves(), 'Feature Importances: ')
print(clf.feature_importances_)
print()

Max Depth:  5
Training Score:  0.7588181176975681 Test Score:  0.7576670435193268
Number of Leaves:  32 Feature Importances: 
[0.00468159 0.17712135 0.03478141 0.00822626 0.17812585 0.01288492
 0.37357755 0.01411648 0.07845248 0.09357435 0.02445777]


# for i in reversed(range(1, 40)):
#     clf = DecisionTreeClassifier(max_depth = i, random_state = 42)
#     clf.fit(X_train, y_train)
#     train = clf.score(X_train, y_train)
#     test = clf.score(test_x, test_y)
#     print('Max Depth: ', clf.get_depth())
#     print('Training Score: ', train, 'Test Score: ', test)
#     print('Number of Leaves: ', clf.get_n_leaves(), 'Feature Importances: ')
#     print(clf.feature_importances_)
#     print()


clf = DecisionTreeClassifier(max_depth = 5, random_state = 42)
clf.fit(X_train, y_train)
#plt the figure, setting a black background
plt.figure(figsize=(30,10), facecolor ='w')
#create the tree plot
a = tree.plot_tree(clf,
                   #use the feature names stored
                   feature_names = x.columns,
                   #use the class names stored
                   class_names = ['Q1', 'Q2', 'Q3', 'Q4'],
                   rounded = True,
                   filled = True,
                   fontsize=14)
#show the plot
plt.show()


x = df.drop(columns=['match_id', 'account_id', 'hero_id', 'player_slot', 'localized_name','cluster', 'region', 'team', 'stay', 'total_wins', 'total_matches', 'trueskill_mu', 'trueskill_sigma', 'mu_bin', 'sigma_bin', 'leave', 'leaver_status'])
y = df["stay"]


X_train, test_x, y_train, test_y = train_test_split(x,y,
                                                 test_size = 0.3,
                                                 random_state = 42)


clf = DecisionTreeClassifier(random_state = 42)
clf.fit(X_train, y_train)

print('Max Depth: ', clf.get_depth())
print('Training Score: ', clf.score(X_train, y_train), 'Test Score: ', clf.score(test_x, test_y))
print('Number of Leaves: ', clf.get_n_leaves(), 'Feature Importances: ')
print(clf.feature_importances_)
print()

Max Depth:  56
Training Score:  1.0 Test Score:  0.5361563890836756
Number of Leaves:  81223 Feature Importances: 
[0.16585109 0.14593648 0.14948426 0.07470016 0.06053611 0.08115638
 0.04614478 0.1672135  0.05296385 0.0560134 ]


# for i in reversed(range(1, 56)):
#     clf = DecisionTreeClassifier(max_depth = i, random_state = 42)
#     clf.fit(X_train, y_train)
#     train = clf.score(X_train, y_train)
#     test = clf.score(test_x, test_y)
#     print('Max Depth: ', clf.get_depth())
#     print('Training Score: ', train, 'Test Score: ', test)
#     print('Number of Leaves: ', clf.get_n_leaves(), 'Feature Importances: ')
#     print(clf.feature_importances_)
#     print()


clf = DecisionTreeClassifier(random_state = 42, max_depth=6)
clf.fit(X_train, y_train)

print('Max Depth: ', clf.get_depth())
print('Training Score: ', clf.score(X_train, y_train), 'Test Score: ', clf.score(test_x, test_y))
print('Number of Leaves: ', clf.get_n_leaves(), 'Feature Importances: ')
print(clf.feature_importances_)
print()

Max Depth:  6
Training Score:  0.6146125302830507 Test Score:  0.6156389083675602
Number of Leaves:  64 Feature Importances: 
[0.01625801 0.02844027 0.07514479 0.         0.15888705 0.0919079
 0.53657526 0.02512881 0.01132047 0.05633744]


clf = DecisionTreeClassifier(max_depth = 6, random_state = 42)
clf.fit(X_train, y_train)
#plt the figure, setting a black background
plt.figure(figsize=(30,10), facecolor ='w')
#create the tree plot
a = tree.plot_tree(clf,
                   #use the feature names stored
                   feature_names = x.columns,
                   #use the class names stored
                   class_names = ['Leave', 'Stay'],
                   rounded = True,
                   filled = True,
                   fontsize=14)
#show the plot
plt.show()


x = df.drop(columns=['match_id', 'account_id', 'hero_id', 'player_slot', 'localized_name','cluster', 'region', 'team', 'stay', 'total_wins', 'total_matches', 'trueskill_mu', 'trueskill_sigma', 'mu_bin', 'sigma_bin', 'leave', 'leaver_status'])
y = df["leaver_status"]


X_train, test_x, y_train, test_y = train_test_split(x,y,
                                                 test_size = 0.3,
                                                 random_state = 42)


clf = DecisionTreeClassifier(random_state = 42)
clf.fit(X_train, y_train)

print('Max Depth: ', clf.get_depth())
print('Training Score: ', clf.score(X_train, y_train), 'Test Score: ', clf.score(test_x, test_y))
print('Number of Leaves: ', clf.get_n_leaves(), 'Feature Importances: ')
print(clf.feature_importances_)
print()

Max Depth:  50
Training Score:  0.9999969333619553 Test Score:  0.9680510039068649
Number of Leaves:  9007 Feature Importances: 
[0.43914928 0.0944293  0.09251337 0.03795894 0.05154526 0.04129124
 0.04664842 0.11388143 0.03314714 0.04943563]


# for i in reversed(range(1, 30)):
#     clf = DecisionTreeClassifier(max_depth = i, random_state = 42)
#     clf.fit(X_train, y_train)
#     train = clf.score(X_train, y_train)
#     test = clf.score(test_x, test_y)
#     print('Max Depth: ', clf.get_depth())
#     print('Training Score: ', train, 'Test Score: ', test)
#     print('Number of Leaves: ', clf.get_n_leaves(), 'Feature Importances: ')
#     print(clf.feature_importances_)
#     print()


clf = DecisionTreeClassifier(random_state = 42, max_depth=6)
clf.fit(X_train, y_train)

print('Max Depth: ', clf.get_depth())
print('Training Score: ', clf.score(X_train, y_train), 'Test Score: ', clf.score(test_x, test_y))
print('Number of Leaves: ', clf.get_n_leaves(), 'Feature Importances: ')
print(clf.feature_importances_)
print()

Max Depth:  6
Training Score:  0.9832070900671593 Test Score:  0.9821400460809708
Number of Leaves:  60 Feature Importances: 
[9.20117058e-01 2.38517371e-02 1.55239441e-02 1.58113923e-03
 4.38987641e-03 7.66946080e-03 1.42565089e-04 2.08522083e-03
 3.50244041e-04 2.42887545e-02]


clf = DecisionTreeClassifier(max_depth = 6, random_state = 42)
clf.fit(X_train, y_train)
#plt the figure, setting a black background
plt.figure(figsize=(30,10), facecolor ='w')
#create the tree plot
a = tree.plot_tree(clf,
                   #use the feature names stored
                   feature_names = x.columns,
                   #use the class names stored
                   class_names = ['0','1','2','3','4','5','6'],
                   rounded = True,
                   filled = True,
                   fontsize=14)
#show the plot
plt.show()


test_pred_decision_tree = clf.predict(test_x)
#get the confusion matrix
confusion_matrix = metrics.confusion_matrix(test_y,  
                                            test_pred_decision_tree)
#turn this into a dataframe
matrix_df = pd.DataFrame(confusion_matrix)
#plot the result
ax = plt.axes()
sns.set(font_scale=1.3)
plt.figure(figsize=(10,7))
sns.heatmap(matrix_df, annot=True, fmt="g", ax=ax, cmap="magma")
#set axis titles
ax.set_title('Confusion Matrix - Decision Tree')
ax.set_xlabel("Predicted label", fontsize =15)
# ax.set_xticklabels(['']+labels)
ax.set_ylabel("True Label", fontsize=15)
ax.set_yticklabels(['0','1','2','3','4'], rotation = 0)
plt.show()

<Figure size 720x504 with 0 Axes>


x = df.drop(columns=['match_id', 'account_id', 'hero_id', 'player_slot', 'localized_name','cluster', 'region', 'team', 'stay', 'total_wins', 'total_matches', 'trueskill_mu', 'trueskill_sigma', 'mu_bin', 'sigma_bin', 'leave', 'leaver_status'])
y = df["leave"]


X_train, test_x, y_train, test_y = train_test_split(x,y,
                                                 test_size = 0.3,
                                                 random_state = 42)


clf = DecisionTreeClassifier(random_state = 42)
clf.fit(X_train, y_train)

print('Max Depth: ', clf.get_depth())
print('Training Score: ', clf.score(X_train, y_train), 'Test Score: ', clf.score(test_x, test_y))
print('Number of Leaves: ', clf.get_n_leaves(), 'Feature Importances: ')
print(clf.feature_importances_)
print()

Max Depth:  34
Training Score:  1.0 Test Score:  0.9886801093349743
Number of Leaves:  2971 Feature Importances: 
[0.43496684 0.1154963  0.09714021 0.03504039 0.04636621 0.04900602
 0.04453715 0.10822572 0.02139042 0.04783073]


# for i in reversed(range(1, 34)):
#     clf = DecisionTreeClassifier(max_depth = i, random_state = 42)
#     clf.fit(X_train, y_train)
#     train = clf.score(X_train, y_train)
#     test = clf.score(test_x, test_y)
#     print('Max Depth: ', clf.get_depth())
#     print('Training Score: ', train, 'Test Score: ', test)
#     print('Number of Leaves: ', clf.get_n_leaves(), 'Feature Importances: ')
#     print(clf.feature_importances_)
#     print()


clf = DecisionTreeClassifier(random_state = 42, max_depth=7)
clf.fit(X_train, y_train)

print('Max Depth: ', clf.get_depth())
print('Training Score: ', clf.score(X_train, y_train), 'Test Score: ', clf.score(test_x, test_y))
print('Number of Leaves: ', clf.get_n_leaves(), 'Feature Importances: ')
print(clf.feature_importances_)
print()

Max Depth:  7
Training Score:  0.9937103253702966 Test Score:  0.9930735435121714
Number of Leaves:  105 Feature Importances: 
[0.7977691  0.07207692 0.04349728 0.00601541 0.00853717 0.00660776
 0.00941537 0.00981136 0.00087897 0.04539067]


clf = DecisionTreeClassifier(max_depth = 7, random_state = 42)
clf.fit(X_train, y_train)
#plt the figure, setting a black background
plt.figure(figsize=(30,10), facecolor ='w')
#create the tree plot
a = tree.plot_tree(clf,
                   #use the feature names stored
                   feature_names = x.columns,
                   #use the class names stored
                   class_names = ['Stay', 'Leave'],
                   rounded = True,
                   filled = True,
                   fontsize=14)
#show the plot
plt.show()


test_pred_decision_tree = clf.predict(test_x)
#get the confusion matrix
confusion_matrix = metrics.confusion_matrix(test_y,  
                                            test_pred_decision_tree)
#turn this into a dataframe
matrix_df = pd.DataFrame(confusion_matrix)
#plot the result
ax = plt.axes()
sns.set(font_scale=1.3)
plt.figure(figsize=(10,7))
sns.heatmap(matrix_df, annot=True, fmt="g", ax=ax, cmap="magma")
#set axis titles
ax.set_title('Confusion Matrix - Decision Tree')
ax.set_xlabel("Predicted label", fontsize =15)
# ax.set_xticklabels(['']+labels)
ax.set_ylabel("True Label", fontsize=15)
ax.set_yticklabels(['stay', 'leave'], rotation = 0)
plt.show()

<Figure size 720x504 with 0 Axes>


from sklearn import linear_model
from sklearn.metrics import mean_squared_error, r2_score


x = df.drop(columns=['match_id', 'account_id', 'hero_id', 'player_slot', 'localized_name','cluster', 'region', 'team', 'stay', 'total_wins', 'total_matches', 'trueskill_mu', 'trueskill_sigma', 'mu_bin', 'sigma_bin'])
y = df['trueskill_mu']
X_train, test_x, y_train, test_y = train_test_split(x,y,
                                                 test_size = 0.3,
                                                 random_state = 42)


regr = linear_model.LinearRegression()
regr.fit(X_train, y_train)
y_pred = regr.predict(test_x)


regr.score(test_x, test_y)

0.0036545430731829143


# Split into a training and testing set
X_train, X_test, y_train, test_y = train_test_split(x, y, random_state=42)

# Define the pipeline for scaling and model fitting
pipeline = Pipeline([
    ("Standard Scaling", StandardScaler()),
    ("Linear Regression", linear_model.LinearRegression())
])

# Scale the data and fit the model
pipeline.fit(X_train, y_train)

# Evaluate the model
y_pred = pipeline.predict(X_test)
print('Mean Absolute Error: ', mean_absolute_error(y_pred, test_y))
print('Score', pipeline.score(X_test, test_y))

Mean Absolute Error:  1.964669725588323
Score 0.0035632551383004696


# The mean squared error
print("Mean squared error: %.2f" % mean_squared_error(test_y, y_pred))
# The coefficient of determination: 1 is perfect prediction
print("Coefficient of determination: %.2f" % r2_score(test_y, y_pred))

Mean squared error: 8.45
Coefficient of determination: 0.00


for i in range(x.shape[1]):
    X_train, X_test, y_train, y_test = train_test_split(x.iloc[:,i].values.reshape(-1,1), y, test_size=0.8, random_state=42)
    # reg = linear_model.LinearRegression().fit(X_train, y_train)
    pipeline = Pipeline([
    ("Standard Scaling", StandardScaler()),
    ("Linear Regression", linear_model.LinearRegression())
])

    reg = pipeline.fit(X_train, y_train)

    y_pred = reg.predict(X_test)

    # The Feature name
    print('Feature: ', x.iloc[:,i].name)
    # The coefficients
    # print("Coefficients: \n", reg.coef_)
    # The mean squared error
    print("Mean squared error: %.2f" % mean_squared_error(y_test, y_pred))
    # The coefficient of determination: 1 is perfect prediction
    print("Coefficient of determination: %.2f" % r2_score(y_test, y_pred))

    plt.plot(X_test, y_pred, color="blue")
    plt.scatter(X_train, y_train, color="red", alpha = .5, linewidths=.001, marker='^')
    plt.scatter(X_test, y_test, color="green", alpha = .5, linewidths=.001)


    # plt.xticks(())
    # plt.yticks(())
    plt.xlabel('X')
    plt.ylabel('y')
    # plt.legend(loc='best')

    plt.title(x.iloc[:,i].name)
    plt.show()

Feature:  gold
Mean squared error: 8.42
Coefficient of determination: 0.00

Feature:  gold_per_min
Mean squared error: 8.41
Coefficient of determination: 0.00

Feature:  xp_per_min
Mean squared error: 8.42
Coefficient of determination: 0.00

Feature:  kills
Mean squared error: 8.42
Coefficient of determination: 0.00

Feature:  deaths
Mean squared error: 8.41
Coefficient of determination: 0.00

Feature:  assists
Mean squared error: 8.42
Coefficient of determination: -0.00

Feature:  denies
Mean squared error: 8.41
Coefficient of determination: 0.00

Feature:  hero_damage
Mean squared error: 8.42
Coefficient of determination: 0.00

Feature:  hero_healing
Mean squared error: 8.42
Coefficient of determination: 0.00

Feature:  level
Mean squared error: 8.42
Coefficient of determination: 0.00


x = df.drop(columns=['match_id', 'account_id', 'hero_id', 'player_slot', 'localized_name','cluster', 'region', 'team', 'stay', 'total_wins', 'total_matches', 'trueskill_mu', 'trueskill_sigma', 'mu_bin', 'sigma_bin', 'leave', 'leaver_status'])
y = df["mu_bin"]

X_train, test_x, y_train, test_y = train_test_split(x,y,
                                                 test_size = 0.3,
                                                 random_state = 42)


logr = linear_model.LogisticRegression()
logr.fit(X_train, y_train)

c:\ALICE_Env\lib\site-packages\sklearn\linear_model\_logistic.py:814: ConvergenceWarning: lbfgs failed to converge (status=1):
STOP: TOTAL NO. of ITERATIONS REACHED LIMIT.

Increase the number of iterations (max_iter) or scale the data as shown in:
    https://scikit-learn.org/stable/modules/preprocessing.html
Please also refer to the documentation for alternative solver options:
    https://scikit-learn.org/stable/modules/linear_model.html#logistic-regression
  n_iter_i = _check_optimize_result(

LogisticRegression()


logr.score(test_x, test_y)

0.7572305622737096


from sklearn.linear_model import SGDRegressor
from sklearn.preprocessing import StandardScaler
from sklearn.preprocessing import MinMaxScaler
from sklearn.pipeline import Pipeline
from sklearn.metrics import mean_absolute_error
from sklearn import linear_model


# Split into a training and testing set
X_train, X_test, Y_train, Y_test = train_test_split(x, y, random_state=42)

# Define the pipeline for scaling and model fitting
pipeline = Pipeline([
    ("Standard Scaling", StandardScaler()),
    ("Logistic Regression", linear_model.LogisticRegression())
])

# Scale the data and fit the model
pipeline.fit(X_train, Y_train)

# Evaluate the model
Y_pred = pipeline.predict(X_test)
print('Mean Absolute Error: ', mean_absolute_error(Y_pred, Y_test))
print('Score', pipeline.score(X_test, Y_test))

Mean Absolute Error:  0.24371248744214802
Score 0.7578502674715141

c:\ALICE_Env\lib\site-packages\sklearn\linear_model\_logistic.py:814: ConvergenceWarning: lbfgs failed to converge (status=1):
STOP: TOTAL NO. of ITERATIONS REACHED LIMIT.

Increase the number of iterations (max_iter) or scale the data as shown in:
    https://scikit-learn.org/stable/modules/preprocessing.html
Please also refer to the documentation for alternative solver options:
    https://scikit-learn.org/stable/modules/linear_model.html#logistic-regression
  n_iter_i = _check_optimize_result(


print('Score', pipeline.score(X_train, Y_train))

Score 0.758651680247751


# Split into a training and testing set
y = df["sigma_bin"]
X_train, X_test, Y_train, Y_test = train_test_split(x, y, random_state=42)

# Define the pipeline for scaling and model fitting
pipeline = Pipeline([
    ("Standard Scaling", StandardScaler()),
    ("Logistic Regression", linear_model.LogisticRegression())
])

# Scale the data and fit the model
pipeline.fit(X_train, Y_train)

# Evaluate the model
Y_pred = pipeline.predict(X_test)
print('Mean Absolute Error: ', mean_absolute_error(Y_pred, Y_test))
print('Score', pipeline.score(X_test, Y_test))

Mean Absolute Error:  0.24371248744214802
Score 0.7578502674715141

c:\ALICE_Env\lib\site-packages\sklearn\linear_model\_logistic.py:814: ConvergenceWarning: lbfgs failed to converge (status=1):
STOP: TOTAL NO. of ITERATIONS REACHED LIMIT.

Increase the number of iterations (max_iter) or scale the data as shown in:
    https://scikit-learn.org/stable/modules/preprocessing.html
Please also refer to the documentation for alternative solver options:
    https://scikit-learn.org/stable/modules/linear_model.html#logistic-regression
  n_iter_i = _check_optimize_result(


print('Score', pipeline.score(X_train, Y_train))

Score 0.758651680247751


# Split into a training and testing set
y = df["stay"]
X_train, X_test, Y_train, Y_test = train_test_split(x, y, random_state=42)

# Define the pipeline for scaling and model fitting
pipeline = Pipeline([
    ("Standard Scaling", StandardScaler()),
    ("Logistic Regression", linear_model.LogisticRegression())
])

# Scale the data and fit the model
pipeline.fit(X_train, Y_train)

# Evaluate the model
Y_pred = pipeline.predict(X_test)
print('Mean Absolute Error: ', mean_absolute_error(Y_pred, Y_test))
print('Score', pipeline.score(X_test, Y_test))

Mean Absolute Error:  0.38462661320098573
Score 0.6153733867990142


print('Score', pipeline.score(X_train, Y_train))

Score 0.6148524684944602


# Split into a training and testing set
y = df["leave"]
X_train, X_test, Y_train, Y_test = train_test_split(x, y, random_state=42)

# Define the pipeline for scaling and model fitting
pipeline = Pipeline([
    ("Standard Scaling", StandardScaler()),
    ("Logistic Regression", linear_model.LogisticRegression())
])

# Scale the data and fit the model
pipeline.fit(X_train, Y_train)

# Evaluate the model
Y_pred = pipeline.predict(X_test)
print('Mean Absolute Error: ', mean_absolute_error(Y_pred, Y_test))
print('Score', pipeline.score(X_test, Y_test))

Mean Absolute Error:  0.008397661019568783
Score 0.9916023389804313


print('Score', pipeline.score(X_train, Y_train))

Score 0.9916595827501624


# Split into a training and testing set
y = df["leaver_status"]
X_train, X_test, Y_train, Y_test = train_test_split(x, y, random_state=42)

# Define the pipeline for scaling and model fitting
pipeline = Pipeline([
    ("Standard Scaling", StandardScaler()),
    ("Logistic Regression", linear_model.LogisticRegression())
])

# Scale the data and fit the model
pipeline.fit(X_train, Y_train)

# Evaluate the model
Y_pred = pipeline.predict(X_test)
print('Mean Absolute Error: ', mean_absolute_error(Y_pred, Y_test))
print('Score', pipeline.score(X_test, Y_test))

Mean Absolute Error:  0.03602064210336507
Score 0.9769364851752947

c:\ALICE_Env\lib\site-packages\sklearn\linear_model\_logistic.py:814: ConvergenceWarning: lbfgs failed to converge (status=1):
STOP: TOTAL NO. of ITERATIONS REACHED LIMIT.

Increase the number of iterations (max_iter) or scale the data as shown in:
    https://scikit-learn.org/stable/modules/preprocessing.html
Please also refer to the documentation for alternative solver options:
    https://scikit-learn.org/stable/modules/linear_model.html#logistic-regression
  n_iter_i = _check_optimize_result(


print('Score', pipeline.score(X_train, Y_train))

Score 0.9771397005578405


from sklearn.decomposition import PCA
from sklearn.decomposition import TruncatedSVD
from sklearn.preprocessing import StandardScaler
from scipy.sparse import csr_matrix

import matplotlib.pyplot as plt
import seaborn as sns


y = df['stay']
scaler = StandardScaler()
scaler.fit(x)
X_scaled = scaler.transform(x)

pca = PCA(n_components=2)
pca.fit(X_scaled)
xpca = pca.transform(X_scaled)

print(pca.explained_variance_ratio_*100)

[44.21475457 14.52720259]


plt.plot(pca.explained_variance_ratio_)

[<matplotlib.lines.Line2D at 0x1e448a014c0>]


plt.figure(figsize=(10,7))
sns.scatterplot(x=xpca[:,0], y=xpca[:,1], hue=y)

<AxesSubplot:>


y = df['leave']
scaler = StandardScaler()
scaler.fit(x)
X_scaled = scaler.transform(x)

pca = PCA(n_components=2)
pca.fit(X_scaled)
xpca = pca.transform(X_scaled)

print(pca.explained_variance_ratio_*100)

[44.21475457 14.52720259]


plt.plot(pca.explained_variance_ratio_)

[<matplotlib.lines.Line2D at 0x1e45b3094c0>]


plt.figure(figsize=(10,7))
sns.scatterplot(x=xpca[:,0], y=xpca[:,1], hue=y)

<AxesSubplot:>


y = df['leaver_status']
scaler = StandardScaler()
scaler.fit(x)
X_scaled = scaler.transform(x)

pca = PCA(n_components=2)
pca.fit(X_scaled)
xpca = pca.transform(X_scaled)

print(pca.explained_variance_ratio_*100)

[44.21475457 14.52720259]


plt.plot(pca.explained_variance_ratio_)

[<matplotlib.lines.Line2D at 0x1e45b42ff70>]


plt.figure(figsize=(10,7))
sns.scatterplot(x=xpca[:,0], y=xpca[:,1], hue=y)

<AxesSubplot:>

c:\ALICE_Env\lib\site-packages\IPython\core\pylabtools.py:137: UserWarning: Creating legend with loc="best" can be slow with large amounts of data.
  fig.canvas.print_figure(bytes_io, **kw)


y = df["mu_bin"]
plt.figure(figsize=(10,7))
sns.scatterplot(x=xpca[:,0], y=xpca[:,1], hue=y)

<AxesSubplot:>


y = df["sigma_bin"]
plt.figure(figsize=(10,7))
sns.scatterplot(x=xpca[:,0], y=xpca[:,1], hue=y)

<AxesSubplot:>


y = df["mu_bin"]
plt.figure(figsize=(10,7))
sns.scatterplot(x=xpca[:,0], y=xpca[:,1], hue=y)

<AxesSubplot:>


df.head()


df.groupby('region')['trueskill_sigma'].mean().plot(legend=True, kind='bar')

<AxesSubplot:xlabel='region'>

	name	hero_id	localized_name
0	npc_dota_hero_antimage	1	Anti-Mage
1	npc_dota_hero_axe	2	Axe
2	npc_dota_hero_bane	3	Bane
3	npc_dota_hero_bloodseeker	4	Bloodseeker
4	npc_dota_hero_crystal_maiden	5	Crystal Maiden

	account_id	total_wins	total_matches	trueskill_mu	trueskill_sigma
0	236579	14	24	27.868035	5.212361
1	-343	1	1	26.544163	8.065475
2	-1217	1	1	26.521103	8.114989
3	-1227	1	1	27.248025	8.092217
4	-1284	0	1	22.931016	8.092224

	account_id	hero_id	player_slot	gold	gold_per_min	xp_per_min	kills	deaths	assists	...	localized_name	total_wins	total_matches	trueskill_mu	trueskill_sigma	cluster	region	stay	team
0	0	86	0	3261	347	362	9	3	18	...	Rubick	1608398	3315071	25.000000	8.333333	155	SINGAPORE	0	0
1	0	83	2	110	350	385	0	4	15	...	Treant Protector	1608398	3315071	25.000000	8.333333	155	SINGAPORE	0	0
2	0	102	129	317	303	369	4	13	5	...	Abaddon	1608398	3315071	25.000000	8.333333	155	SINGAPORE	0	1
3	0	7	131	475	189	223	1	14	8	...	Earthshaker	1608398	3315071	25.000000	8.333333	155	SINGAPORE	0	1
4	1	51	1	2954	494	659	13	3	18	...	Clockwerk	14	24	26.232905	4.854238	155	SINGAPORE	1	0

	match_id	account_id	hero_id	player_slot	gold	gold_per_min	xp_per_min	kills	deaths	assists	...	hero_healing	level	leaver_status	leave	total_wins	total_matches	trueskill_mu	trueskill_sigma	cluster	team
stay
0	24949.026904	0.000000	49.260259	65.613997	1800.868317	407.682726	447.708924	7.436775	8.030006	11.645295	...	428.004835	18.124259	0.044192	0.012221	1.608398e+06	3.315071e+06	25.00000	8.333333	142.335832	0.49692
1	24948.018572	59398.434576	51.468958	66.207716	1953.339621	421.308439	452.371650	7.400812	7.433898	11.816378	...	490.510914	18.008908	0.034558	0.007605	1.920866e+01	3.664298e+01	25.81447	5.425989	140.927610	0.50170

	match_id	account_id	hero_id	player_slot	gold	gold_per_min	xp_per_min	kills	deaths	assists	...	hero_healing	level	leaver_status	total_wins	total_matches	trueskill_mu	trueskill_sigma	cluster	stay	team
leave
0	24953.209127	36344.710177	50.610085	65.960279	1908.547257	417.468950	452.543528	7.453822	7.672920	11.804009	...	469.148592	18.109289	0.01425	623720.194992	1.285549e+06	25.499862	6.551333	141.466568	0.612218	0.499711
1	24442.746061	31480.582325	50.602877	67.720484	365.785796	262.214661	241.358986	3.302352	6.906828	6.042019	...	155.785567	12.202786	2.57319	813205.538479	1.676098e+06	25.273107	7.114394	142.390500	0.494405	0.513588

Association¶

Functions¶

Code¶

Match Analysis¶

FP¶

Functions¶

Decision Tree¶

mu¶

Sigma¶

Stay¶

Status¶

Leave¶

Linear Regression¶

Plots¶

Logistic Regression¶

Mu¶

Sigma¶

Stay¶

Leave¶

Status¶

PCA¶

Stay¶

Leave¶

Status¶

Mu¶

	account_id	hero_id	player_slot	gold	gold_per_min	xp_per_min	kills	deaths	assists	denies	hero_damage	hero_healing	level
0	0	86	0	3261	347	362	9	3	18	1	8690	218	16
1	1	51	1	2954	494	659	13	3	18	9	23747	0	22
2	0	83	2	110	350	385	0	4	15	1	4217	1595	17
3	2	11	3	1179	599	605	8	4	19	6	14832	2714	21
4	3	67	4	3307	613	762	20	3	17	13	33740	243	24

	match_id	account_id	hero_id	player_slot	gold	gold_per_min	xp_per_min	kills	deaths	assists	...	hero_healing	level	leave	total_wins	total_matches	trueskill_mu	trueskill_sigma	cluster	stay	team
leaver_status
0	24963.978047	36327.702978	50.609736	65.912365	1924.295219	418.587029	454.184817	7.496398	7.670898	11.876585	...	472.881690	18.166874	0.0	624438.148236	1.287028e+06	25.499015	6.553147	141.462328	0.611771	0.499339
1	24208.278589	37521.168187	50.634276	69.274635	819.195864	340.126825	339.008820	4.508668	7.812804	6.783607	...	210.914842	14.125912	0.0	574056.412561	1.183186e+06	25.558471	6.425823	141.759884	0.643096	0.525395
2	24277.188320	30922.121879	50.277190	65.903936	429.704190	281.392298	266.725349	3.878967	7.166737	7.321202	...	189.637749	13.466356	1.0	872611.296234	1.798539e+06	25.218060	7.301820	144.178163	0.457469	0.499365
3	24712.184625	34009.176084	50.476347	69.863338	20.871222	230.065703	198.344941	2.173456	6.400788	3.762812	...	78.084757	9.883706	1.0	668944.119580	1.378760e+06	25.404191	6.731906	140.109724	0.584100	0.530223
4	24404.542510	26361.396761	52.550607	69.807692	1122.710526	269.530364	252.546559	4.022267	7.222672	6.945344	...	233.251012	13.303644	1.0	973509.331984	2.006500e+06	25.132549	7.396302	140.866397	0.394737	0.530364

	match_id	account_id	hero_id	player_slot	gold	gold_per_min	xp_per_min	kills	deaths	assists	...	hero_healing	level	leaver_status	leave	total_wins	total_matches	trueskill_mu	trueskill_sigma	cluster	stay
team
0	24950.699167	36144.263026	50.503022	1.997283	1901.857461	418.775975	451.780679	7.407303	7.672660	11.761743	...	460.350843	18.068533	0.036739	0.009142	629154.583856	1.296749e+06	25.496566	6.568899	141.502777	0.608839
1	24946.120783	36453.808682	50.717082	129.997032	1886.227732	413.241342	449.335270	7.422297	7.658773	11.737941	...	472.058746	18.038991	0.039872	0.009659	621845.862601	1.281685e+06	25.498897	6.544344	141.447713	0.613383

	cluster	region
0	111	US WEST
1	112	US WEST
2	113	US WEST
3	121	US EAST
4	122	US EAST

	match_id	cluster
0	0	155
1	1	154
2	2	132
3	3	191
4	4	156